(1317) Craniofacial Imaging Predictors of Hypoglossal Nerve Stimulation Response: A Systematic Review and Meta-analysis

Introduction: Hypoglossal nerve stimulation (HNS) is an established sleep surgery implant for CPAP-intolerant obstructive sleep apnea (OSA), yet ~1/3 of patients do not respond. We performed a systematic review and meta-analysis of imaging-derived anatomic predictors of HNS outcomes.

Methods: A PRISMA-guided review identified adult OSA patients treated with hypoglossal nerve stimulation (HNS) who underwent imaging-based anatomic assessment, including cephalometry/lateral radiographs, computed tomography (CT) or four-dimensional CT (4DCT), ultrasound-based tongue analysis, and endoscopy-derived grading using awake endoscopy and/or drug-induced sleep endoscopy (DISE). Random-effects meta-analyses pooled mean apnea–hypopnea index (AHI) change, with standard deviations of change derived assuming within-subject correlation (r = 0.50) when unreported, and pooled surgical response rates using a logit model. Response was defined as ≥50% AHI reduction with postoperative AHI < 15 or < 20 events/hour per original study definitions; heterogeneity was assessed using I².

Results: Sixteen studies (n = 738) met inclusion criteria. Seven studies (n = 133) contributed extractable apnea–hypopnea index (AHI) data; pooled mean AHI reduction following hypoglossal nerve stimulation was −23.9 events/hour (95% CI −30.7 to −17.1; I² = 85.8%). Eight studies (n = 605) reported Sher-style response outcomes; pooled responder proportion was 67.7% (95% CI 57.9%–76.2%; I² = 84.4%). Across modalities, cephalometric markers of craniofacial retrusion (smaller SNA/SNB angles, increased mandibular–hyoid distance), CT-based soft palate enlargement, and vertical palate configuration were consistently associated with non-response, while ultrasound-derived tongue tissue composition and protrusion mechanics on dynamic imaging were associated with treatment response.

Conclusions: Imaging phenotyping reveals convergent anatomic signals associated with HNS response and supports imaging-informed implant candidacy. Standardized metrics and prospective validation are needed to define actionable thresholds.